QSAR models

Key biological mechanisms such as the activation of nuclear receptors, the inhibition of enzyme activity or the opening of ion channels can be perturbed by biological active chemicals. Because of these molecular interactions, the maintenance of cellular homeostasis can be disrupted and adverse outcomes can occur at an organism level (e.g. hepatic steatosis, impaired fertility).

The (eco)toxicological potential of the large number of existing and newly synthesized chemicals to which living organisms are exposed can be carried out by means of in vitro and in vivo experiments and by means of in silico (computational models) screening. The latter approach offer the advantages of being much less time-consuming and of not being associated to ethical issue related to animal testing.

Among the available tools, Quantitative Structure Activity Relationships (QSAR models) play a special role since the toxicological predictions that can be estimated by these relationships can be used in regulatory contexts.

The underlying hypothesis of QSAR models is that “similar molecules induce similar biological effects”. During the development of QSAR models, this working hypothesis is characterized by computational algorithms (e.g. linear or non-linear regressions) that formalize correlations between a toxicological endpoint (e.g. mutagenicity) and the chemical structures of a set of molecules of known toxicity. During this analysis chemicals structure are described by numerical variables which are generally referred to as “molecular descriptors”. These descriptors are used to quantify the physicochemical, topological, geometrical and quantum mechanical properties of the chemical of interest.

Several international initiatives and projects promote the use of QSAR models in regulatory context and several “QSAR toolboxes” are freely available to all the stakeholders interested in regulations on chemical hazards. QSAR tools can therefore be easily used alone or together with other existing information in order to enable a sound decision-making in the field of regulatory toxicology or other research areas that rely on the characterization of structure-activity relationships (e.g. drug design).

Key publications

• • Mombelli, E., Ringeissen S. 2009. The computational prediction of toxicological effects in regulatory contexts. Act. Chim. 335:52-59.

  • Tebby C., Mombelli E. 2012. A Kernel-Based Method for Assessing Uncertainty on Individual QSAR Predictions. Mol Inform. 31:741-51.

  • Mombelli E. 2012 Evaluation of the OECD (Q)SAR Application Toolbox for the profiling of estrogen receptor binding affinities. SAR QSAR Environ Res. 23:37-57.

  • Mombelli E., Raitano G., Benfenati E. 2016. In Silico Prediction of Chemically Induced Mutagenicity: How to Use QSAR Models and Interpret Their Results. Methods Mol Biol. 1425:87-105.